3783aeab64
# What this PR does - updates the GitHub Actions workflow to move the e2e tests into a "[reusable workflow](https://docs.github.com/en/actions/using-workflows/reusing-workflows#creating-a-reusable-workflow)" which are run in two scenarios: - all tests _except_ those annotated as `@expensive` are run against `grafana/grafana:latest` on all feature branches - all tests _including_ `@expensive` tests are run on weekdays @ 07h00 UTC, against a matrix of 6 grafana versions. Results of these builds will be posted to `#irm-amixr-flux` Slack channel. - local development will now be: ```bash make build-dev-images init-k8s start-k8s ``` - `build-dev-images` - builds the engine and UI docker images (only need to run first time) - `init-k8s` - creates a `kind` cluster and loads the two Docker images onto the cluster nodes (only need to run first time) - `start-k8s` - switches `kubectl` context to the created `kind` cluster, and uses `helm` to deploy everything as defined in `./dev/helm-local.yml` and `./dev/helm-local.dev.yml` (that latter file is `.gitignored` and specific to how _you_ want your setup to look like. Hot reloading works as before. This is the _start_ of #2381. (I've marked these `make` commands as beta, because they've not yet been thoroughly tested for local development). - modifies the `helm` chart to add the concept of `oncall.devMode`, `ui`, and ability to run oncall w/ sqlite - `oncall.devMode` will essentially just add `volumes` and `volumeMounts` to the various engine/migrate containers + - `ui.enabled` + `ui.env` - create a ui container (which is needed for hot reloading locally) - `sqlite` - this was useful for the e2e test environments where Github runner resources are scarce. Running `mariadb` eats up precious resources, instead lets just use sqlite here - fixes an issue that caused sporadic HTTP 502s from the grafana plugin-proxy, which led to flaky tests. See [this comment](https://github.com/grafana/oncall/pull/2751/files#diff-09040e8df192699b9c5742110ebbe8d9d5c3938cb156cc1cb99fa1c3fdee4fefR72-R77) for more context + a link to a relevant Slack conversation. **tldr;** there is a bug with the Grafana plugin proxy in Grafana >= v10.0.3. Let's stop using the `latest`/`main` docker tags in our test and pin to `10.0.2` for now - ~~re-enables the e2e test which validates a phone number via SMS, and asserts that we can receive an alert escalation via SMS (new Mailslurp API Key has been added as a repo secret)~~ update: this is still blocked by procurement, will be done in a future PR ## Checklist - [x] Unit, integration, and e2e (if applicable) tests updated - [x] Documentation added (or `pr:no public docs` PR label added if not required) - [x] `CHANGELOG.md` updated (or `pr:no changelog` PR label added if not required) |
||
|---|---|---|
| .. | ||
| oncall | ||
| cr.yaml | ||
| ct.yaml | ||
| kind.yml | ||
| README.md | ||
| simple.yml | ||
How to run the chart locally
-
Create the cluster with kind
Make sure ports 30001 and 30002 are free on your machine
kind create cluster --image kindest/node:v1.24.7 --config kind.yml -
(Optional) Build oncall image locally and load it to kind cluster
-
docker build ../engine -t oncall/engine:latest --target dev kind load docker-image oncall/engine:latest -
Install the helm chart
helm install helm-testing \ --wait \ --values ./simple.yml \ ./oncall -
Get credentials
echo "\n\nOpen Grafana on localhost:30002 with credentials - user: admin, password: $(kubectl get secret --namespace default helm-testing-grafana -o jsonpath="{.data.admin-password}" | base64 --decode ; echo)" echo "Open Plugins -> Grafana OnCall -> fill form: backend url: http://host.docker.internal:30001" -
Clean up
kind delete cluster